Validity of the Driver Behavior Questionnaire

8/12/2019 Validity of the Driver Behavior Questionnaire

1/9

Accident Analysis and Prevention 52 (2013) 228236

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier .com/ locate /aap

Age, gender, mileage and the DBQ: The validity ofthe Driver BehaviorQuestionnaire in different driver groups

Laila M. Martinussen a,, Liisa Hakamies-Blomqvistb, Mette Mllera, Trker zkanc, Timo Lajunenc,d

a DTUTransport, Bygningstorvet 115, DK-2800 Kgs, Lyngby, Denmarkb Culture and Society, Academy of Finland, Hakaniemenranta 6, POB 131, FI-00531 Helsinki, Finlandc Department of Psychology, Middle East Technical University (ODT), Inonu Bulvari, 06531 Ankara, Turkeyd Department of Psychology, NorwegianUniversity of Science andTechnology (NTNU), 7491 Trondheim, Norway

a r t i c l e i n f o

Article history:Received 16 November 2011Received in revised form11 December 2012Accepted 13 December 2012

Keywords:Driver Behavior Questionnaire (DBQ)Factor structureGenderAgeMileage

a b s t r a c t

The Driver Behavior Questionnaire (DBQ) is one ofthe most widely used instruments for measuring self-reported driving behaviors. Despite the popularity ofthe DBQ, the applicability ofthe DBQin differentdriver groups has remained mostly unexamined. The present study measured aberrant driving behaviorusing the original DBQ(Reason, J.T., Manstead, A., Stradling, S.G., Baxter, J., Campbell, K., 1990. Errorsand violations on the road a real distinction. Ergonomics, 33 (10/11), 13151332) to test the factorialvalidity and reliability ofthe instrument across different subgroups ofDanish drivers. The survey wasconducted among 11,004 Danish driving license holders ofwhom 2250 male and 2190 female driverscompletedthe questionnaire containing background variables and the DBQ. Exploratory and confirmatoryfactor analysis showed that the original three-factor solution, a four-factor solution and a two-factorsolution had acceptable fit when using the whole sample. However, fit indices ofthese solutions variedacross subgroups. The presents study illustrates that both the original DBQ and a Danish four-factorDBQ structure is relatively stable across subgroups, indicating factorial validity and reliability of theDBQ. However, as the Danish DBQstructure has an overall better fit, the present study highlights theimportance ofperforming an explorative analysis when applying the DBQin order to assess the problem

areas within a driving population. 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The classification of behavioral items in the Driver BehaviorQuestionnaire (DBQ) is based on Reasons theory, namely genericerror modeling system (GEMS) (Reason, 1990). The original DBQwas designed and developed by Reason, Manstead, Stradling, Bax-ter, and Campbell (1990) to measure aberrant driving behaviorwith 50 items measuring lapses, errors and violations. Since then,it has become one of the most widely used instruments for mea-suring both driving style (Bener et al., 2006) and the relationship

between driving behavior andcrash involvement (for a review see:de Winter and Dodou, 2010).

The DBQ has over the years been applied in numerous countriesfor example; Qatar and United Arab Emirates (Bener et al., 2008),USA (Owsley et al., 2003), China (Xie and Parker, 2002), Australia(Blockey and Hartley, 1995; Davey et al., 2007; Dobson et al., 1999;Lawton et al., 1997), Sweden (Rimm and Hakamies-Blomqvist,2002; berg and Rimm, 1998; berg and Warner, 2008), Greece

Corresponding author. Tel.: +45 45 25 65 00; fax: +45 45 93 65 33.E-mail address: [email protected](L.M. Martinussen).

(Kontogiannis et al., 2002), The Netherlands (Lajunen et al., 1999),Spain (Gras et al., 2006), France (Obriot-Claudel and Gabaude,2004), New Zealand (Sullman et al., 2000), Turkey (zkan andLajunen, 2005; Smer, 2003), and UK (Parker et al., 1995; Reasonet al., 1990). However, the factorial structures of the DBQ as wellas the numberof items vary between different driving cultures andnations.

Many studies have found support for the original three-factorstructure consisting oflapses, errors and violations (Dobson et al.,1999; Kontogiannis et al. , 2002; Reason et al., 1990; berg and

Rimm, 1998). Others have found that aggressive violations, ordi-nary violations and lapses were applicable, although not firmlystable across countries (Walln Warner et al., 2011). Similar resultshave been obtained in Australia by Lawton et al. (1997) and Daveyet al.(2007) who foundsupport for errors, highway-code violationsand interpersonal aggressive violations. However, also within Aus-tralian drivers Blockey and Hartley (1995) found a different factorstructure consisting of general errors, dangerous errors and dan-gerous violations in their DBQ study.

In addition to the content of the factors, the number of factorshas also varied between studies; Hennessy and Wiesenthal (2005)and Smer (2003) reported fewer factors and Kontogiannis et al.

0001-4575/$ see front matter 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.aap.2012.12.036
http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.aap.2012.12.036http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.aap.2012.12.036http://www.sciencedirect.com/science/journal/00014575http://www.elsevier.com/locate/aapmailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.aap.2012.12.036http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.aap.2012.12.036mailto:[email protected]://www.elsevier.com/locate/aaphttp://www.sciencedirect.com/science/journal/00014575http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.aap.2012.12.036


2/9

L.M. Martinussen et al. / Accident Analysis andPrevention 52 (2013) 228236 229

(2002), and Parker et al. (2000) more factors than in the originalDBQ, which might partly reflect the number and different itemcontents. The most commonbesides the original three-factor struc-ture seems to be the four-factor solution (Mesken et al. , 2002;Lajunen et al., 2004; Rimm, 2002; Xie and Parker, 2002). Despitecross-cultural differences, the importantdistinction between unin-tended errors and intended violations has been found in moststudies (Kontogiannis et al. , 2002; Blockey and Hartley, 1995;Lajunen et al., 2004; Parker et al., 1998; Rimm and Hakamies-Blomqvist, 2002; Sullman et al., 2002; zkan et al., 2006a; WallnWarner, 2006). The distinction between errors and violations alsoseems to be stable over time (zkan et al., 2006b). Moreover, theliterature also reports variations in driving style among subgroupssuch as age, genderand annualmileage (Lawtonet al., 1997;Reasonet al., 1990; Rimm, 2002; Rimm and Hakamies-Blomqvist, 2002;berg and Rimm, 1998).

Despite the popularity of the DBQ (de Winter and Dodou, 2010,reports 174 studies using some version of the DBQ), no studyso far has tested the fit of the original DBQ model across driversubgroups. Only two studies have employed confirmatory factoranalysis (CFA) to test the factorial validity of the DBQ (Rimm,2002; zkan et al., 2006a). zkan et al. (2006a) used CFA to testthe applicability of a three-factor model (aggressive violations,

ordinary violations and errors) across six countries. Rimm (2002)investigated the fit of the Swedish DBQ (DBQ-SWE) across differ-ent driver subgroups: new drivers, inexperienced drivers, youngdrivers and experienced drivers. However, Rimm focused mainlyon young drivers anddid notmake any distinction between driversagedfrom28to70.Inaddition,theDBQ-SWEincludesonly32itemsfrom Reasonet al.s original 50 item DBQ. It would thereforebe per-tinent to test the fit of the original DBQ in different drivers groups,as Reason et al.s DBQ is the original from which all other versionshave been derived, and also because it has been suggested that dif-ferent driver subgroups could best be tested with different DBQversions (Rimm, 2002).

The first aim of the present study was to investigate if thedistinction between errors and violations were present in the

sample of Danish drivers as this structure seems to be themost stable across studies. The second aim was to develop acountry specific Danish DBQ which could be used in furtherstudies of aberrant driver behavior in Denmark. The third aimwas to investigate the applicability of the three different DBQstructures(the two-factorstructure, the original three-factor struc-ture and the Danish factor structure) among different driversubgroups.

2. Methods

2.1. Participants and procedure

Drivers with a type B driver license (Danish license for per-sonal car) were randomly selected from the Danish Driving LicenseRegister. The sample was stratified by age and gender to include1572 drivers in each of the following seven age groups; 1824years, 2534 years, 3544 years, 4554 years, 5564 years, 6574years, 7584 years (respectively 786 men and women in eachage group). The questionnaire together with a cover letter and afreepost return envelope were sent by post to all 11,004 selectedparticipants. A web address that the respondents could use toreply was also included. Two reminders were sent. The totalresponse rate was 44 percent. Of the 4849 responses, 4335 per-sons had fully completed the DBQ. Participants responded to thequestionnaire anonymously. The Danish Data Protection Agencyhad approved the survey. Sample characteristics can be found inTable 1.

Table 1

Sample characteristics.

Total Males Females

n 4335 2204 2131AgeMean 50.9 53.25 48.5St.D. 18.886 19.049 18.416Annual mileage (km)Mean 16,251.56 20,204.88 11,971.41

St.D. 28,401.28 29,001.27 27,100.97

2.2. Measures

The DBQ and demographic measures were combined into onequestionnaire as part of a larger study. Respondents answeredquestions about age and gender, as well as last years annual driv-ing distance. The original Driver Behavior Questionnaire (Reasonet al., 1990) was translated into Danish using the back-translationmethod. The drivers were asked, using the standard DBQ instruc-tions (see Reason et al., 1990), to indicate on a six-point Likert scale(0=never and 5= nearly all the time) how often they performedeach of the 50 driving behaviors. Since Reason et al. (1990) only

reported items which had factor loadings above 0.50, only 27 ofthe original 50 items were in thecurrent study used as the originalDBQ.

2.3. Statistical analysis

Exploratory factor analysis (EFA, principal axis with obliminrotation) and confirmatory factor analyses (CFA, LISREL with max-imum likelihood estimation) were performed in order to examinethe underlying dimensions and the model fit (see Russell, 2002for detailed information regarding confirmatory and exploratoryanalysis). In the EFA, scree plots, interpretability of the factors, andparallel analysis were used to determine the number of factors tobe extracted as the Danish DBQ. In addition, an EFA with a forced

two-factor solution was performed. A CFA was carried out in orderto examine the fit of the model established in the EFA, the simplertwo-factor model, as well as the original DBQ (1990) structure inthewholesample andacross subgroups. In line with HuandBentler(1999) and Bryne (2001) the fit of the models was evaluated by2/degree of freedom ratio, root mean square error of approxima-tion (RMSEA), comparative fit index (CFI) and standardized rootmean square residual (SRMR). A good fit model should have 2:1 or5:1 2/degree of freedom ratio, CFI > 0.90 (preferably > 0.95), andRMSEA


3/9


4/9


F1Unfocusederrors/

lapses

F3Recklessviolations/

lapses

F2Emotionalviolations

F4Confusederrors/

lapses

DBQ11(.56)

DBQ12(.45)

DBQ20(.50)

DBQ24(.55)

DBQ30(.57)

DBQ25

(.52)

DBQ28(.57)

DBQ32(.52)

DBQ36(.42)

DBQ49

(.46)

DBQ46(.62)

DBQ42(.52)

DBQ41(.49)

DBQ40(.44)

DBQ43(.35)

DBQ44 (.46)

DBQ47(.56)

DBQ9(.55)

DBQ50(.53)

DBQ7

(.59)

DBQ18(.40)

DBQ19(.36)

DBQ27(.41)

DBQ29(.43)

DBQ35(.42)

DBQ4(.52)

DBQ2(.46)

DBQ37(.52)

DBQ45(.68)

DBQ23(.37)

DBQ21(.69)

DBQ5(.44)

DBQ15(.45)

DBQ14(.51)

DBQ13(.44)

DBQ8(.46)

DBQ38(.49)

DBQ10(.42)

DBQ33

(.51)

DBQ17(.50)

.47

.51

.36

.43

.76

DBQ48

(.50)

.69

.69

.73

.70

.79

.75

.82

.68

.73

.68

.80

.79

.61

.76

.70

.72

.88

.79

.69

.72

.66

.84

.87

.83

.81

.82

.73

.79

.54

.86

.53

.81

.73

.80

.74

.80

.78

.76

.82

.74

.75

.68

Fig. 1. DBQ structurewith four factors. The figure shows factorloadings and error measures for all items.

3.2. Fit of the three models

CFA were performed in order to test the fit of the original DBQmodel, the Danish four-factor model revealed in the EFA, and theforced two-factor model. The two-,three and four-factor structuresused in the analysis are schematically presentedin Figs. 13, which

respectively show two-, three and four factors that inter-correlateto explain aberrant driver behavior. No items loaded on more thanone factor. The goodness of fit indices suggest satisfactory, butnot perfect fit for all three structures in the whole sample, with aslightly betterfit for thethree-and four-factor solutions (two-factorsolution: CFI 0.828, RMSEA 0.043, SRMR 0.045, 2/df 7032.75/778;three-factor solution: CFI 0.848, RMSEA 0.045, SRMR 0.043; 2/df3197.76/321; four-factor solution; CFI 0.848, RMSEA 0.040; SRMR0.046, 2/df 6207.81/773; see Table 3).

Furthermore, the three DBQ models were applied to the dataconsisting of different driver subgroups (see Table 3). Results sug-gest that the three- and the four-factor models had a reasonablygoodfit amongolder drivers (men andwomen analyzed separately)aswellas a goodfit in all annual mileage groups. The three- and thefour-factor models had the poorest fit among the younger drivers.

The two-factor model was generally less fitting than the two otherfactor models. The two-factor model had the poorest fit amongyoung and middle-aged men and women. Fit indices are presentedin Table 3 (correlation matrixes can be obtained upon request fromthe corresponding author). Looking at Table 3, one can see thatthe three-factor model had the best fit across sub-groups of drivers

accordingtotheCFIstatistics,however,whenlookingattheRMSEA,the four-factor model had the best fit across groups. In general, andin all models, it was better fit among older than younger drivers.

3.3. Factor interpretability in both EFA and CFA

In theDanish four-factor solution,mainly error andlapses itemsloaded on the first factor, which seems to contain an underlyingstructure of actions performed while unfocused; thus it can benamed unfocused errors/lapses. The second factor can be labeledemotional violations as the loading items were violations trig-geredby emotionalarousal.The thirdfactor included violationsandlapses. The underlying structureseems to be reckless behavior,andcan thus be called reckless violations/lapses. In the fourth factor,

the items represent errors and lapses characterized by confusion,


5/9

232 L.M. Martinussen et al. / Accident Analysis andPrevention 52 (2013) 228236

F1

Errors

F2

Violations

DBQ9

(.56)

DBQ10

(.37)

DBQ11(.55)

DBQ12

(.47)

DBQ17

(.40)

DBQ14

(.44)

DBQ15(.37)

DBQ20

(.49)

DBQ23

(.41)

DBQ31(.42)

DBQ30

(.56)

DBQ28

(.56)

DBQ25(.51)

DBQ24

(.53)

DBQ27(.40)

DBQ29 (.42)

DBQ35(.39)

DBQ8

(.38)

DBQ32

(.51)

DBQ5

(.37)

DBQ7(.57)

DBQ16

(.52)

DBQ18

(.37)

DBQ19(.32)

DBQ21

(.61)

DBQ4

(.51)

.48

DBQ39

(.48)

.86

.69

.81

.78

.86

.69

.83

.84

.76

.86

.72

.83

.69

.74

.69

.74

.84

.82

.85

.77

.87

.67

.73

.87

.90

.82

.74

DBQ36

(.41)

DBQ37(.46)

DBQ38

(.43)

DBQ40(.43)

DBQ46(.60)

DBQ41(.49)

DBQ42

(.51)

DBQ49(.46)

DBQ50(.51)

DBQ33(.45).80

..83

.76

.81

.79

.81

.74

.64

.78

.74

DBQ45

(.56)

DBQ47(.55)

DBQ48

(.51)

DBQ44

(.40)

.69

.70

.74

.84

Fig. 2. DBQ structure with two factors. The figure shows factor loadings and error measures forall items.

Table 3

Fit indexes from confirmatory factor analysis.

2 Factors 3 Factors 4 Factors

CFI RMSEA 2(df778) Ratio SRMR CFI RMSEA 2(df321) Ratio SRMR CFI RMSEA 2(df773) Ra tio SRMR

Whole sample (n = 4335) .828 .043 7032.75 9.04 .045 .848 .045 3197.76 9.96 .043 .848 .040 6207.81 8.03 .046GenderMen (n= 2204) .839 .042 3843.63 4.94 .045 .867 .046 1810.15 5.69 .044 .851 .040 3538.82 4.58 .048

Women (n = 2131) .804 .045 4079.38 5.24 .047 .854 .046 1757.95 5.48 .044 .840 .041 3479.15 4.50 .045AgeYoung 1829 (n= 779) .782 .047 2099.51 2.67 .058 .817 .053 1010.91 3.15 .058 .804 .044 1954.42 2.53 .056Middle 3049 (n= 1336) .788 .047 3038.55 3.91 .053 .832 .049 1336.44 4.16 .049 .811 .044 2779.09 3.60 .052Old 5085 (n= 2220) .844 .041 3668.25 4.72 .044 .883 .041 1522.98 4.74 .041 .859 .039 3334.71 4.31 .045Gender*AgeYoung men 1829 (n= 327) .749 .051 1449.72 1.86 .069 .776 .059 690.21 2.15 .070 .768 .049 1375.09 1.78 .067Middle men 3049 (n= 649) .767 .048 1932.32 2.48 .056 .813 .050 843.11 2.63 .054 .784 .046 1848.61 2.37 .056Old men 5085 (n= 1228) .850 .041 2418.91 3.11 .047 .880 .043 1041.89 3.25 .046 .863 .048 2217.81 2.87 .045Young woman 1829 (n= 452) .697 .055 1831.94 2.36 .065 .768 .058 807.47 2.52 .065 .754 .050 1629.04 2.11 .062Middle women 3049(n= 687) .784 .047 1970.58 2.53 .055 .822 .050 876.83 2.73 .052 .813 .044 1818.78 2.35 .055Old women 5085 (n= 992) .795 .045 2351.35 3.02 .050 .862 .043 899.50 2.80 .045 .813 .043 2208.04 2.86 .051Annual driving distanceLow (16558 km) (n= 1258) .809 .044 2691.03 3.46 .047 .867 .044 1104.20 3.44 .045 .845 .039 2286.93 2.96 .045Middle (655915,000km) (n= 1060) .813 .046 2531.38 3.25 .050 .860 .046 1055.24 3.29 .046 .832 .043 2305.64 2.98 .051High (15,00110,5000 km) (n = 1317) .801 .046 2917.85 2.96 .051 .841 .049 1314.91 4.1 .050 .810 .045 2816.10 3.64 .053

Note: criteriafora goodfit are 2:1 or5:12/df, CFI> 0.90, RMSEA< 0.05,and SRMR< 0.08.


6/9


Violations

Errors

Lapses

DBQ45(.60)

DBQ44(.38)

DBQ21(.64) DBQ14(.55)

DBQ15(.47)

DBQ17(.47)

DBQ28(.57)

DBQ20(.51)

DBQ11(.58)

.46

.73

.31

.64

.59

.85

.70

.78

.78

.68

.74

.66

DBQ30(.59)

DBQ

16

(.53)

DBQ18(.35)

DBQ19(.34)

DBQ

46

(.63)

DBQ42(.53)

DBQ41(.52)

DBQ32(.52)

DBQ5(.41)

DBQ7(.55)

DBQ48(.49)

DBQ47(.59)DBQ37(.56)

DBQ10(.44)

DBQ2(.40)

DBQ4(.50)

DBQ38(.54)

DBQ8(.43)

.81

.81

.69

.71

.88

.88

.72

.69

.83

.75

.84

.76

.66

.66

.73

.73

.72

.60

Fig. 3. DBQ structure with three factors. The figure shows factor loadings and error measures forall items.

and can therefore be calledconfusederror/lapses (see Table 2 andFig. 1).

In the two-factor model, factor one contained mostly unin-tended errors and lapses and can therefore be labeled as errors.The second factor contained a mix of emotional and ordinary vio-lations and could be labeled violations (see Table 2 and Fig. 2).

3.4. Inter-correlations and reliability analysis in four- andtwo-factor solutions

Correlations between the twoviolation factors and between thetwothe lapse factors were higherthan between anyof theviolationfactors and lapse factors.

Factor one, i.e., unfocused error/lapses, showed the highestinternal consistency, whereas the factor three, reckless viola-tions/lapses, had the lowest internal consistency. Factor twoand four, emotional violations and confused error/lapses, respec-tively, both had acceptably high alpha values, 0.730 and 0.724(Cortina, 1993), showing good internal consistency. Alpha valuesare also in line with the original study and other previous stud-

ies (Lajunen et al., 2004; zkan et al., 2006a; Reason et al., 1990).

Inter-correlations and alpha values for all three factor models canbe seen in Table 4.

4. Discussion

Firstly, the EFA revealed a distinction between error/lapses andviolations, thus clearly showing the difference between intendedand unintended aberrant driving behavior. Secondly, the EFA

revealed a Danish DBQ structure consisting of four factors: twoerror/lapses factors named confused- and unfocused errors/lapses,one emotional violation factor and one reckless violation/lapsesfactor. Further, the fit of the original DBQ, the Danish DBQ, and theforced two-factor DBQ structure was tested with CFA. Acceptablefit was found for both the original DBQ three-factor model and thefour-factor model in the whole sample. Lastly, fit of the three mod-elsweretestedacrosssubgroupsdifferingingender,ageandannualmileage. The original three-factor modeland the Danish four-factormodel had the best fit across all subgroups compared to the two-factor solution. However, in the whole sample, the older sample aswell as in gender groups separately, the fit of the two-factor modelcouldbe considered acceptable.In general, thefit wasbetter amongtheolderdrivers compared to theyoungerones.The present results

show validity and reliability of the original DBQ model, as well as


7/9


Table 4

Inter-correlations (Pearsons r) and alpha value between theDBQ factors.

Factors UL EV RV CL Alphavalues

Factors V E L Alphavalues

Factors E V Alphavalues

UL 1 .302 .280 .578 .850 V 1 .358 .255 .735 E 1 .391 .875EV .302 1 .318 .132 .730 E .358 1 .516 .769 V .391 1 .770RV .280 .318 1 .230 .637 L .255 .516 1 .679CL .578 .132 .230 1 .724

Note: UL= unfocused errors/lapses, EV= emotional violations, RV= reckless violations/lapses, CL= confused lapses.V = violations, E =errors, L =lapses.

for the Danish DBQ model, thus supporting the further use of bothmodels.

4.1. The two-factormodel

The distinction between intended behavior (violations) andunintended behavior (errors and lapses) was salient in the Danishpopulation, as lapses and errors items loaded together and viola-tion items together. This distinction was expected due to earlierfindings (see Walln Warner, 2006 p. 2728 for an overview) andthe different psychological processes behind lapses and errors, andviolations, as highlighted by Reason et al. (1990). Violations are

motivational and/or contextual based, while errors are cognitivebased. The difference of the two behavioral classes was validatedusing CFA. This further supports the theory behind DBQ, that aber-rant driving behavior can be separated into two broad behavioralclasses of unintentional errors/lapses and intentional violations.

4.2. The DanishDBQ factor model

In Reason et al. (1990), a cut-off point of 0.50 for the factorloadings was applied. In the present study, the cut-off point of0.50 was found too high resulting in many deleted items and lowinterpretability of factors, so a lower (0.30) cut-off point wasapplied (see Costello and Osborn, 2005; Field, 2009; Kline, 1994)and a different factorial solution than presented in Reason et al.

(1990) was revealed.There are important differences between the principal compo-

nent analysis (PCA) used by Reason et al. (1990) and the principalaxis factoring (PAF) used in the present study. Different factorextraction methods (PCA vs. PAF), and rotation techniques (vari-max vs. oblimin) may explain the differences between the resultsofReason et al. (1990) and the present study. When applying PCA,loadings become higher than in PAF because of higher communal-ity estimates (Russell, 2002). The PCA has been the most commonmethod in the DBQ literature; which is somewhat peculiar sincethe literature in the field of factor analysis generally recommendsPAF over PCA (Reise et al., 2000; Russell, 2002; Widaman, 1993).

The current EFA revealed a four-factor structure containing twofactors explained by both error and lapses items and two factors

containing violations items. The factors could be said to resem-ble other studies four-factor solutions. The unfocused lapses/errorsfactor resembles Rimms (2002) mistake factor, Lajunen et al.s(2004) errorsfactorand Meskenetal.s(2002) errorsfactor.Further,the emotional violations factor resembles Mesken et al.s (2002)interpersonal violations and Lajunen et al.s (2004) aggressiveviolations. The present reckless violations/lapses factor consistsof both violations and lapses, but does not resemble previousfactor solutions in the literature. The confused lapses/errors resem-ble Rimms (2002) inattention errors factor and Mesken et al.s(2002) lapses factor. However, the present factor structure doesnot seem to separate between errors and lapses, as both behav-iors load together. Thus, the distinction between errors and lapsesis not present in the Danish sample. The implication of this

is that the broad distinction of behavioral classes in the DBQ

between errors/lapses and violations, thus intended versus unin-tended behavior, is further supported. Additionally, the distinctionbetween the aberrant behavioral classes is not stable, as differ-ent underlying structures do seem to appear when applying theDBQ in different countries. Originally, the DBQ was thought to con-sist of five factors or behavioral classes (mistakes, lapses, slips,unintended violations and deliberate violations). However, Reasonet al.s study (1990) did not find such a structure, but found a three-factor structure instead. Other previous studies have also founddifferent factor structures of the DBQ (Blockey and Hartley, 1995;Lawton et al., 1997; Reason et al., 1990; berg and Rimm, 1998).This is not surprising as the driving style is formed by personalfactors such as age, gender and cognitive biases, as well as by thesocial context (Reason, 1990; Reason et al., 1990). The fact that thepresent four-factor structure resembles previously obtained factorstructures, although not completely, confirms the need to applyexplorative analysis when the DBQ is applied in a population withthe purpose to identify relevant preventive efforts. The differentfactors found are indicative of the relevant preventive strategy.Drivers who perform many emotional violations need other meansto change their driving style than drivers performing reckless driv-ing violations as there is different underlying mechanism anddifferent motivational mechanisms behind. Further, errors andlapses caused by confusion are different than errors/lapses causedby voluntary engagement in distracting behaviors. The currentstudy, as well as previous studies (Walln Warner et al., 2011),find that different countries have differentproblems with regardto

aberrant driving behavior. A country or populations factor struc-ture is a good indicator of where preventive efforts should betargeted.

4.3. Fit of the three DBQ models

Results showed that the original three-factor structure and theDanish four-factor structure had an acceptable fit in the wholesample whereas the forced two-factor structure showed a some-what lower fit. This could reflect the complexity of the driver tasks,thus a more complex model explains driver behavior to a greaterdegree. On the other hand, the difference between the fit indexesof the three models was small, indicative of stable DBQ structures

across driver groups. Overall a slightly better fit was obtained bythe Danish four-factor structure than by the original three-factorstructure.

Further, the fit of the three DBQ models was tested across sub-groups. The three-factor model had the best fit across sub-groupsof drivers according to the CFI statistics, however when lookingat the RMSEA, the four-factor model had the best fit across groups.Since theCFI statistics in general arebelow therecommendedvalue(good fit> 0.90) and the RMSEA statistics are in the recommendedend (good fit < 0.05) across sub-groups, and that the literature doesnot recommend one over the other, the four-factor model seemsslightly better fitting. This is not surprising as the EFA did notreveal a three-factor structure, thus the four-factor structure rep-resents the presentsample better. As for the whole population, the

three- and four-factor models revealed a slightly better fit than the


8/9


9/9


Obriot-Claudel, F., Gabaude, C., 2004. The driver behaviour questionnaire: a Frenchstudy applied to elderly drivers. In: Paper presented at the 3rd InternationalConference on Traffic & Transport Psychology, Nottingham, UK, September.

Owsley, C., McGwin Jr., G., McNeal, S.F., 2003. Impact of impulsiveness, venture-someness, and empathy on driving by older adults. Journal of Safety Research34, 353359.

zkan, T., Lajunen, T., 2005. A new addition to DBQ: positive driver behavior scale.Transportation Part F: Traffic Psychologyand Behaviour 8 (4-5), 355368.

zkan,T., Lajunen, T.,Chliaoutakis, J.,Parker, D.,Summala,H., 2006a.Cross-culturaldifferences in driving behviours: a comparison of six countries. TransportationPart F: Traffic Psychology and Behaviour 9 (3), 227242.

zkan, T.,Lajunen,T., Summala,H., 2006b.Driverbehaviourquestionnaire:a follow-up study. Accident Analysis and Prevention 38, 386395.

Parker, D., Lajunen, T., Stradling,S.G., 1998. Attitudinalpredictors of interpersonallyaggressive violations on theroad. Transportation Part F: Traffic PsychologyandBehaviour 1 (1), 1124.

Parker, D., McDonald, L., Rabbit, P., Sutcliffe, P., 2000. Elderly drivers and their acci-dents: theaging driver questionnaire. Accident Analysis and Prevention 32 (6),751759.

Parker, D., Reason, J.T., Manstead, A., Stradling, S.G., 1995. Driver errors, drivingviolations, and accident involvement. Ergonomics 38 (5), 10361048.

Reason, J.T., 1990. Human Errors. Cambridge University Press, Cambridge.Reason, J.T., Manstead, A., Stradling, S.G., Baxter, J., Campbell, K., 1990. Errors and

violations on the road a real distinction. Ergonomics 33 (10/11), 13151332.Reise, S.P., Waller, N.G., Comrey, A.L., 2000. Factor analysis and scale revision. Psy-

chological Assessment 12 (3), 287297.Rimm, P.-A., 2002.Aberrantdriving behaviour:homogeneityof a four-factor struc-

ture in samples differing in age and gender. Ergonomics 4 (8), 569582.

Rimm, P.-A., Hakamies-Blomqvist, L., 2002. Older drivers aberrant drivingbehaviour impaired activity, and health as reasons for self-imposed driving lim-itations. Transportation Research Part F: TrafficPsychology and Behaviour 5 (1),4762.

Russell, D.W.,2002.In search of underlyingdimensions:the use(and abuse)of factoranalysis in Personality and Social Psychological Bulletin (2002). Personality andSocial Psychological Bulletin 28, 16291646.

Sullman, M., Meadows, M., Pajo, K., 2000. Traffic and Transport Psychology. Theoryand Application. Elsevier, Amsterdam.

Sullman, M., Meadows, M., Pajo,K., 2002.AberrantdrivingbehavioursamongstNewZealand truck drivers. Transportation Research Part F: Traffic Psychology and

Behaviour 5 (3), 217232.Smer,N., 2003. Personality andbehavioral predictors of trafficaccidents: testing a

contextual mediated model. Accident Analysis and Prevention 35 (6), 949964.Walln W.H., 2006. Factors influencing drivers speeding behaviour. Doctoral dis-

sertation. Acta Universitatis Upsaliensis Digital comprehensive summaries ofUppsaladissertations fromthe Faculty ofSocial Sciences,p. 21.Uppsala,Sweden.(accessed 07.06.11).

Walln, W.H., zkan, T., Lajunen, T., Tzamalouka, G., 2011. Cross-cultural com-parison of drivers tendency to commit different aberrant driving behaviours.Transportation Research Part F 14, 390399.

Widaman, K.F., 1993. Common factor analysis versus principal component analy-sis: differential bias in representing model parameters? Multivariate BehavioralResearch 28 (3), 263311.

Xie, C-qiu., Parker, D., 2002. A social psychological approach to driving violationsin twoChinese cities. Transportation Part F: Traffic Psychology andBehaviour 5(4), 293308.
http://uu.diva-portal.org/smash/record.jsf?searchId=1&%20pid=diva2:169263http://uu.diva-portal.org/smash/record.jsf?searchId=1&%20pid=diva2:169263

Documents

Validity of the Driver Behavior Questionnaire